Targeted therapy of cancer requires a clear understanding of the genetic alterations that drive malignant cell growth. Identification of causal genetic alterations is complicated by three characteristics of cancer etiology: 1.) multiple interacting alterations are often required to cause cancer, 2.) several distinct alterations may be sufficient to generate a single cancer phenotype, and 3.) oncogenic alterations appear in a dense background of normal genetic activity and spurious consequences of malignant cell growth. We propose to apply a variant of the machine learning algorithm PRIM to the task of identifying disjunctive sets of conjunctive genetic alterations that cause specific cancers or provide prognostic information about clinical course and treatment efficacy. These analyses synthesize information from low-level bioinformatics resources we have already developed to map chromosomal alterations and monitor global patterns of transcription factor activity. Based on those foundations, the present studies develop high-level analytic tools to map combinatorial interactions among low-level genomic events. Specifically, these studies seek to: Aim 1: Develop graphical user interface (GUI) software to support combinatorial genomic analyses by biologists with limited computational background. Aim 2: Optimize combinatorial prediction of disease progression and treatment response. Aim 3: Develop PRIM-based statistical models to identify functional complementation groups of genetic alterations and transcriptional control signals. The bioinformatic tools produced in these studies will create a generalized analytic infrastructure for mapping complex etiologies in cancer and deploying patient-specific targeted therapies. [unreadable] [unreadable] [unreadable]